How to Check Your Claude Usage Limit (3 Ways)

You're Deep in a Coding Session. Claude Just Stops.

No warning. No countdown timer. No "you have 3 messages left" banner. One moment you're debugging a tricky race condition with full context, the next you're staring at "Usage limit reached." Your conversation history is still there, but Claude won't respond until the window resets.

This is a common frustration among Claude Pro subscribers, and it happens more often than you'd expect. Every time it does, you lose the context you've built up - the files Claude has read, the patterns it's recognized, the approach you've been iterating on together.

Here are three ways to check your usage before that happens.

Method 1: Check claude.ai/settings/usage Manually

The built-in option. Log into claude.ai, click your profile icon, go to Settings, then Usage.

You'll see:

• Your current plan (Pro, Max, Team, or Enterprise)
• Usage across different models (Opus, Sonnet, Haiku)
• How much of your five hour session limit you have used, and the time left in that session
• When your weekly limits reset, listed separately for Opus and for all other models

The problem? You have to leave your current conversation to check it. If you're in the middle of a complex task, switching tabs breaks your flow. And there's no way to set alerts - you'd have to remember to check manually every 30 minutes or so.

Anthropic's usage page also shows aggregate data rather than real-time message counts. The numbers update with a slight delay, so what you see might be a few messages behind your actual usage.

Still, it's free, it's built in, and it requires zero setup. For casual users who aren't pushing their limits daily, this works fine.

Method 2: Use a Menu Bar App

Menu bar apps put your usage data one glance away. No tab switching, no manual checking. You look at the top of your screen and see a number or a progress indicator.

There are several options here (I covered them all in my full usage tracker comparison), but the general idea is the same: the app polls your Claude usage data at set intervals and displays it persistently.

The free options (ClaudeMeter, Usage4Claude, Claude-Usage-Tracker) are macOS-only and show your current usage without notifications. They're great for visual monitoring if you remember to glance up.

OhNine adds threshold notifications - it alerts you at 80%, 91%, and 100% usage via native OS notifications. It runs on macOS 14 or later and also tells you when a session resets. The 91% threshold is the namesake: "oh nine" percent left, "oh nein" in German. It is free.

The key advantage of any menu bar app over the manual method: you see your usage passively. You don't have to decide to check it. It's just there, like your battery indicator or clock.

Method 3: Track via Claude Code CLI

If you're a Claude Code user, you can check usage without leaving your terminal. Claude Code shares the same usage pool as claude.ai - every message counts against the same limits.

Within a Claude Code session, you can:

• Type /usage. On Pro and Max it opens your plan usage bars, with activity stats underneath (/cost is an alias for the same screen)
• Put the limit in your statusline. Claude Code passes rate_limits.five_hour.used_percentage and rate_limits.seven_day.used_percentage to statusline scripts on Pro and Max, so both numbers sit under every prompt. My free Statusline Builder has both as widgets, reset time included.
• Use the --model flag to switch to Haiku when you want to conserve your Sonnet/Opus budget

One catch. The activity breakdown under the bars is built from session history on this machine, so chats on claude.ai or on another laptop are missing from it. For the whole picture across devices, the settings page or a menu bar app is still the reference.

Power users often combine methods 2 and 3: a menu bar app for the big picture, CLI usage tracking for per-session awareness.

Checking Usage on the Free Plan

Free accounts get the same kind of cap: a session limit that resets every five hours. Anthropic does not publish a message count for it and says the number varies with demand.

When you reach it, Claude tells you. Before that, open Settings > Usage on the web or in Claude Desktop, the page Anthropic's help center points to for usage and reset times.

Claude Code is not part of the free plan. Anthropic's sign-in options are Pro, Max, Team, Enterprise or an API key, so /usage and the statusline trick above need a paid plan. If you keep running into the free cap, my Pro vs Max breakdown shows what each tier buys you.

Understanding Claude's Limit System

Claude's usage limits aren't as simple as "X messages per day." Here's how they actually work:

Session Limits (Reset Every 5 Hours)

Claude uses a sliding window, not a fixed daily cap. Your available messages refill over a rolling period of five hours. This means if you burn through your allocation at 9 AM, you'll start getting messages back around 2 PM - but it's gradual, not all at once.

Weekly Limits (Two Separate Counters)

The five hour session window is not the only cap. Paid plans also carry a weekly limit, and it is split in two: one counter for Opus, and one for every other model. That is why you can be locked out of Opus while Sonnet still answers. The usage screen lists both reset times separately, so check the one that matches the model you actually need.

Model-Specific Limits

Not all models cost the same against your limits. Claude Opus messages consume roughly 5x more capacity than Claude Sonnet messages. Haiku messages consume the least. A Pro subscriber might get approximately 45 Opus messages, 225+ Sonnet messages, or thousands of Haiku messages in a given window.

How Limits Differ by Plan

Plan	Monthly Cost	Approx. Sonnet Messages	Approx. Opus Messages	Overflow Option
Pro	20 USD	~225 per window	~45 per window	Pay-per-use
Max 5x	100 USD	~1,125 per window	~225 per window	Pay-per-use
Max 20x	200 USD	~4,500 per window	~900 per window	Pay-per-use
Team	30 USD/seat	Higher than Pro	Higher than Pro	Admin-configured

Anthropic does not publish exact message counts, so treat the two message columns as working estimates. What you actually get swings with conversation length, attachments and how much context each turn carries.

Important: these numbers are approximate. Anthropic adjusts them based on demand, conversation length, and other factors. The exact limits are intentionally not published as hard numbers.

Extra Usage (Pay-Per-Use Overflow)

If you enable "Extra Usage" in your billing settings, Claude keeps working after you hit your plan's limit - but you pay per message at API rates. This can add up fast: heavy users have reported 50-100 USD in overflow charges in a single month. Monitor this carefully if you enable it.

Tips to Stretch Your Usage

Use Haiku for Simple Tasks

Not every question needs Opus or Sonnet. Quick factual lookups, simple formatting, code explanation, and basic debugging work perfectly well on Haiku - and it barely touches your usage limits. Save your heavy models for complex reasoning, architecture decisions, and multi-file refactoring.

Batch Your Complex Work

Instead of sending 20 small messages to Opus, plan your request and send one detailed message. Include all the context Claude needs upfront: the files, the requirements, the constraints. One well-crafted message with a comprehensive response uses less capacity than a back-and-forth conversation that takes 10 turns to reach the same outcome.

Monitor Your Pace

If your window resets every five hours, that gives you one or two fresh windows across a normal workday. Divide your message budget accordingly. If you have 225 Sonnet messages per window, that's about 56 messages per hour. Knowing your pace helps you avoid the "burned through everything by lunch" problem.

Front-Load Context

The biggest waste of messages is re-explaining things Claude already knew in a previous conversation. Use your project's CLAUDE.md file to persist important context. Use Claude Code skills to encode recurring workflows. The less setup each session needs, the more of your budget goes to actual productive work.

Frequently Asked Questions

How often do Claude limits reset?

Claude uses a sliding window that refreshes over five hours. It's not a hard reset at a fixed time - your capacity gradually refills. If you used 100% of your allocation at noon, you'll start seeing capacity come back through the afternoon, with the window clear again around 5 PM. Weekly limits run on their own schedule, with one reset for Opus and another for all other models.

What's the difference between session and weekly limits?

Session limits are the sliding window (five hours) that determines how many messages you can send in a given period. Weekly limits are an additional cap on total usage across an entire week. Most users hit session limits long before weekly limits become relevant. Max subscribers with 5x or 20x allocations are more likely to encounter weekly limits since their per-session capacity is much higher.

Can I pay for more Claude usage?

Yes, in two ways. First, you can upgrade your plan - from Pro (20 USD) to Max 5x (100 USD) or Max 20x (200 USD) for proportionally higher limits. Second, you can enable "Extra Usage" in your billing settings, which lets Claude keep working past your plan's limits at pay-per-use API rates. The second option has no cap, so watch your spend. A menu bar tracker like OhNine helps here - it shows when you've crossed from plan usage into overflow territory.

Does Claude Code use the same limits as claude.ai?

Yes. Claude Code (the CLI coding agent) and the claude.ai web interface share the same usage pool. A message sent in Claude Code counts against the same limits as a message sent on the website. If you use both throughout the day, you're drawing from one shared bucket - which makes tracking even more important.

Can I check my Claude usage on the free plan?

Partly. The free plan has a session limit that resets every five hours, and Claude tells you when you hit it. Settings > Usage on the web or in Claude Desktop is where Anthropic points for usage and reset times. Claude Code and its /usage screen need a paid plan.